migration and the canadian urban system: …...figures 1. the canadian urban system 2. migration...
TRANSCRIPT
MIGRATION AND THE CANADIAN URBAN SYSTEM:
PART III:
COMPARING 1966-1971 and 1971-1976
J.W. SIMMONS*
Research Paper No. 112
*Department of Geography
Centre for Urban and Community Studies University of Toronto
November 1979 Reprinted July 1980
ISSN 0316-0068 ISBN 0-7727-1271-9
Abstract
This paper is the third in a series devoted to the exploration of
interurban migration patterns in Canada. It compares the 1971-1976 patterns
with the 1966-1971 migration described in the first two, looking at both
the migration rates of cities and the flows among them. The predominant
impression is one of considerable stability in the patterns over time,
although the net flows have shifted markedly towards smaller, poorer
centres. The final section develops the implications of these patterns
for population forecasting.
Preface
Within less than five years our knowledge of the local variation in
mobility pattern in Canada has jumped enormously; from virtually zero, to
a point where we can now examine the stability of observed patterns through
time. This paper compares the results :'"rom an analysis of the 1971-1976
patterns with the results of the 1966-1971 data as presented in Research
Papers #85 and U98 in this series. The results suggest a surprising
stability in thase migration rates which have le.d to population grmvth
and thus encourage further efforts to elucidate the forces behind the
patterns. The pattern of migration flows, between pairs of places, in
contrast, demonstrates a strong tende.ncy for increased movement from
large places to smaller centres, while moves from smaller places to larger
places are declining.
This paper is part of an ongoing investigation of growth and change
in the Canadian Urban System (see also Research Papers 62, 65, 70, 93 and
104 in this series) which has been supported in part by a grant from
Canada Council. I would also like to acknowledge the intellectual and
material support of Larry Bourne and John Hitchcock at the Centre for
Urban and Community Studies. Bev Thompson and Cathy Morrissey typed drafts
of the manuscript. Siegfried Schulte helped with the data tapes. Jane
Davies drafted the illustrations.
Jim Simmons September 1979
Contents
The Data Set
Migration Rates Out, In and Net Migration Net Immigration and Natural Increase
Migration Flows The Flow Patterns The Relationships Changes in Flows
Implications for Population Growth Migration and Population Growth The Three Sources of Population Growth Forecasting
Conclusions
References
Appendices A. The Canadian Urban System B. Definition of Variables C. A Note on Flow Relationships for Subsamples of Observations
•
Tables
1. Correlations among Migration Rates
2. Comparing Regression Models of Migration Rates
3. Migration Rates and Change Variables
4. Changes in Migration Rates and Other Variables
5. Regression Models of Changes in Migration Rates
6. Correlations with Natural Increase and Immigration
7. Comparing Regression Models of Natural Increase and Net Immigration
8. Correlations of Natural Increase and Net Immigration with Measures of Economic Growth
9. Changes in Natural Increase and Net Immigration
10. Correlations among Flow Measures over Time
11. Comparing Flow Matrix Models over Time
12. Net Migration Flows and Changes in Economic Conditions
13. Regression Models of Changes in Flow Ratios
14. Interurban Mobility Rates by Order and Region
15. Net Migration among Orders in the Urban Hierarchy
16. Net Migration among Urban Subsystems
17. Components of Population Growth, by Order and Region
18. Variation in Components of Population Growth
19. Regularity of Urban Growth by Order and Subsystem
Figures
1. The Canadian Urban System
2. Migration Status, 1971-1976
3. Net Migration Rate, 1971-1976
4. Changes in Net Migration Rate
5. Size Migrant·Flow, 1971-1976
6. :the Largest Outflow, 1971-1976
7. Net Migration Flows, 1971-1976
8. Net Migration among Urban Subsystems
9. Path Analysis: Population Growth Rate, 1971-1976
10. Path Analysis: Change in Population Growth Rate, 1966-1971 to 1971-1976
11. Components of Population Growth: Canada, 1951-1976
12. Distribution of Components of Population Growth
13. Distribution of Changes in Components of Population Growth
MIGRATION AND THE CANADIAN URBAN SYSTEM: PART III, COMPARING 1966-1971 and 1971-1976 PATTERNS
Virtually every population forecast makes the assumption that
the pattern of net migration will remain essentially unchanged over time.
This assumption is seldom challenged: we know far less about most
aspects of migration than we do about fertility and economic growth,
and almost no studies focus on patterns of migration through time. Yet
those temporal studies that have been done - mainly on net migration pat-
terns of large spatial aggregates - point out that migration patterns
fluctuate rapidly through time, often reversing direction within a de-
cade; and that the patterns, even in one cross-section, are not easily
explained by conventional models. Notable Canadian studies include the
long-run summaries by Kalbach (1970) and Stone (1969) of immigration and
interprovincial migration, respectively; and the study of annual varia-
tions in interprovincial flows by Vanderkamp (1968) and Termote and Frech-
ette (1979). In each case the variability of migration in time and space
is emphasized.
This paper focuses on a smaller spatial unit, the urban centred re-
gion, and makes an explicit comparison of migration patterns from two
time periods, 1966 to 1971 and 1971 to 1976. The 1966-1971 patterns have
been described in detail in two earlier papers (Simmons, 1977 and 1978a) .*
After an introductory discussion of data the paper turns to a comparison
of basic spatial distributions of migration rates, followed by an evalua-
tion of the flow matrices themselves. The final section discusses the
role of migration within the urban growth process as a whole and makes
some suggestions for population forecasting models.
*Hereafter these two pages are referred to as Part I (the 1966-71 patterns) and Part II (1966-71 relationships) respectively.
- 2 -
The Data Set
In both the 1971 and 1976 Census of Canada Statistics Canada asked
a one-third sample of population over five years of age where they had
lived five years before. The responses were coded to the level of county/
census division, generating a matrix of approximately 260 x 260 elements
for each time period. In the analysis to be discussed here these counties/
census divisions were aggregated into 124 urban regions, each centred on
a census metropolitan area, census agglomeration or urban municipality of
at least 10,000 persons. These regions are depicted in Figure 1 and de
scribed in Appendix A. In addition an urban hierarchy has been imposed on
the urban regions to facilitate certain further spatial aggregations of
migration patterns.
The overall composition of migration activity in Canada for 1971 to
1976 is shown in Figure 2, with 1966-1971 values included beneath for
comparison. A slight increase in overall mobility has occurred through
time, due to the increased proportion of population in the high mobility
age-groups - 15 to 30 years. The level of international migration and
the level of intercity migration have both declined, however, reflecting
a slackening in Canadian economic growth.(Bourne and Simmons, 1979).
Note that intercity migrants, as defined in this paper, account for eleven
per cent of all moves.
The Census materials describe the level of migration among urban re
gions, and the rate of immigration into Canada. They do not tell us about
the level of natural increase or the rate of emigration. The latter can
' ' ' '
' ' "
- 3 -
\
\
' "
\ \ \
\ \
\ \
C...'-'-\ )
.. ... :! .. 8 G ~ ..
0 • "
8 0 .. II
~ 2 ~
0 0
u
- 4 -
FIGURE 2
MIGRATION STATUS 1971-1976
1976 Population (5 years and over)
21,239,000 (100.0%)
[19,717,000 (100.0%)]
I Non-Mojers (Some Dwelling) Movers (Different Dwelling)
10,930,000 (51.5%)
[10,371,000 (52.6%)]
J Same Urban Region
7,057,000 (33.2%)
[6,280,000 (31.9%)]
10,309,000
[9,346,000
(48.5%)
(47.4%)]
I
Different Urban Region
r . IntercJ.ty
3,252,000 (15.3%}
[3,0,6,000 (15.5:)]
Abroad
2,532,000
[2,242,000
1966-1971 values are printed beneath
(11.7%)
(11. 3%)]
720,000 (3.4%)
[824,000 (4.2%)]
- 5 -
be estimated; since national figures on population growth (Statistics
Canada, 91-001) indicate that the contribution of net immigration to
population growth during the 1971-1976 period, was 490,000 or 68.0 per
cent of the total immigration. For the 1966-1971 period natural
increase figures were provided for counties and census divisions by
Statistics Canada (91-513), allowing net immigration to be estimated as
follows:
Population Growth = Natural Increase + Net Migration + Net Immigration (1)
Since the number of immigrants is given in the census, the number of
emigrants can be calculated:
Net Immigration = Immigrants from Abroad - Emigrants
and this pattern of emigration was scaled to reflect the temporal change
and then applied to the 1971-1976 pattern. This permitted the calculation
of net immigration during the latter period and then natural increase as
a residual.
MIGRATION RATES
The analysis of migration can be approached through the examination
of the matrix of flows itself, as in the next section; or from the point of
view of its effects on the set of urban places, as discussed below. For
each urban place the sum of out and in migration, hence net (in minus out)
and gross (in plus out) migrants, is given. These measures of growth
and change can be supplemented by information on net immigrants and the
level of natural increase. In each case the measure is converted to a
rate by dividing by the 'base population':
Base Population = (1971 Population + 1976 Population over 5 Years of Age)/2
(2)
(3)
- 6 -
TABLE 1
CORRELATIONS AMONG MIGRATION RATES
Variable 1
1. Out Migration Rate 1.000 1976
2
2. In Migration Rate .639 1.000 1976
3 4 5 6
3. Net Migration Rate 1976
.114 .837 1.000
4. Gross Migration Rate .840 .954 . 636 1. 000 1976
5. Out Migration Rate . 847 .502 .047 . 684 1. 000 1971
6. In Migration Rate .755 .923 .655 . 945 .553 1.000 1971
7 8
7. Net Migration Rate .24CI .no .756 .601 -.119 . 761 1. 000 1971
8. Gross Migration Rate .892 .853 .468 .949 .823 .929 .466 1. 000 1971
9 10
9. Log10 1971 Population
-.435 -.354 -.148 -.419 -.472 -.313 -.006 -.424 1.000
10. Population Growth Rate 1971-1976
Mean
Standard Deviation
Coefficient of Variation
*Absolute Value
.295 .826 .857 .698 .106 .740 .824- .522 -.043 1.000
14.8 16.2 4.8* 31.0 14.2 14.2 4.8* 28.6 4.969 7.0
5.0 9.1 7.1 12.9 4.9 8.3 7.1 11.0 .405 8.8
0.338 0.562 1.48 0.42 0.352 0.585 1.48 0.39 0.082 1.26
- 7 h
Out, In and Net Migration
Table 1 summarizes the correlations between the 1971-1976 migration
rates and those from the previous five years. Note first that in migration
and out migration are again positively correlated (r=.639) with each other,
and that this largely reflects city size variations. Larger places, with
more internal opportunities,send out and attract disproportionately fewer
migrants. The rate of net migration is largely attributable to variations
in the rate of in migration.
The first appraisal of the 1966-1971 data (Part I) concluded that
the most important source of variance at the urban system level was the
presence of two different migration regimes in Canada. In that part of the
country lying east of Montreal, virtually every kind of migration rate had
a low value, while in the rest of the country migration rates were up to
twice as large and with greater variance. "The poorest, most disadvantaged,
areas of the country are least able to adapt quickly to economic change".
This pattern continues, although slightly abated. It is reflected in the
strong temporal stability of the four migration rates considered here.
Out, in, net and gross migration patterns in 1971-1976 correlate with
the previous five year patterns at r=0.847, r=0.923, r=0.7569
and r=0.949
respectively; remarkably high values for this kind of spatial analysis.
In fact, the correlations of out and in migration rates with those of
the previous period are so high that there is no need to produce another
set of maps (see Part I). Figure 3 shows the most recent distribution of
net migration rates, however, in which the familiar patterns of negative
rates in the eastern prairies and Quebec and high net migration in the
Toronto urban field, Alberta and B.C. continue, although the Maritime
provinces are now doing better th~n before.
/
/
// /
I )
<. ' J
( ?
"' \
- \ ( I ; ' )
c-)'~ ; ' \ ' \ '-, I
'- '-..._
' ' ' I '-../.
'-'-,
I I
I I
I
PERCENT NET MIGRATION
1971-1976
-Over 5%
- 1-5% []] -I -+1 °/o
EJ -4--1%
C]up--4%
I
}'.___ I .___
I '
16
14
12
10
8
6
4
2
0
-'
I I
I I
I I
'r---1
I I
T --- --I I I I I I I I I I
Per cent
(8]] Cities east of
Figure 3: Net Migration Rate, 1971....;1976
co
100 200 300 o\00 ~00 MILES
............ -··-----Q IQQ 200 3()0 4QQ "'""' r\II..V-~\no:;~
- 9 -
TABLE 2
Comparing Regression Models of Migration Rates (Beta Coefficients)
1966-71 1971-76
Out Migration Proportion Age 15-34 0.242 0.541
Proportion French Mother Tongue -0.466 -0.644
Log10 Population -o.374 -0.485
Proportion Employment Primary 0.326 0.157
R2 0.488 0.558
In ~figration \'!age Rate 0.511 o. 393.
Employment Ratio 0.301 0.298
Log10
Population -0.538 -0.500
Proportion Mining or -0.259 -o. 291 Manufacturing
R2 0.559 0.420
Net Migration Wage Rate 0.255 -0.063
Employment Ratio 0.175 0.239
Log10 Population -0.364 -0.362
Proportion Primary -0.352 -0.361
Mean January Temperature 0.239 0.279 R2 0.390 0.277
- 10 -
The underlying stability of the migration rates is further confirmed
by Table 2 in which some simple regression models display similar results
for both time periods. These models summarize a considerable experimenta
tion with measures of cross-sectional economic conditions (wages, unem
ployment, economic base) and social characteristics (age, ethnicity,
length of settlement), in which (see Part II) the former performed less
well than expected while the latter were relatively more effective.
Termote and Frechette (1979) present similar findings.
The Table suggests that the model for the out migration rate has
improved due to stronger relationships with the social determinants of
age and ethnicity, but the effectiveness of the in migration rate model
has declined, as the economic variables of wage rates and levels of em
ployment appear to be less important. The level of explanation of net
migration rate has also declined, as the wage rate variable has actually
reversed sign.
Although migration rates are measures of change over a finite per
iod, the independent variables used to explain them in the Part II study
were all cross-sectional variables, measured in 1971. This was done main
ly because of data limitations, but also to simplify the conceptualization.
Nonetheless it is· possible, as Greenwood (1975b) has shown, to relate
migration to other kinds of change which are occurring simultaneously
within the urban system. In one sense this broadens the possible scope
of the study enormously: migration is part of many different kinds of
social, demographic and economic processes, in which it can play both
dependent and independent roles. But in another sense the analysis can
be restricted to those variables in which the spatial structure of change
is a) large in magnitude, relative to the pattern in the first cross-
- 11 -
section ; b) alters the spatial pattern significantly; and c) has a clearly
identifiable structure of its own. Thus the change in the distribution
of the population with French mother tongue is not large, and the changes
in age· structure are largely represented in earlier age structure patterns.
It is the economic measures, in particular level of employment, and to a
lesser extent changes in income or wage level which are the most volatile
in space and time.
Table 3 presents the correlations between the migration rates for
1971-1976 and these change variables, and then evaluates the effectiveness
of the latter in a series of regressions, comparable to Table 2. The re
sults support certain aspects of the economic growth argument. For in
stance, in migration and net migration are strongly associated with em
ployment growth, but out migration is as well and in the same direction,
thougn weakly. While improvements in the ratio of employment to population
of working age and in the wage level help to retard out migration they
also reduce the level of in and net migration. The regression models point
out that the employment growth variable complements the cross-sectional
models in each instance, and markedly improves the performance of the in
migration and net migration models (compare with Table 2). In fact the
ability to create new jobs is. the strongest factor in explaining the net
migration rate ·- which is reassuring. Wage 'improvement, in contrast, is
important in retarding out migration but not in attracting in migrants.
Are these relationships caused in the fashion suggest?d above? Pro
bably only in part - as Muth (1971) and Greenwood (1975a) have suggested.
In growing areas more jobs and more migrants are both found. The unravel
ling of this relationship requires a more sophisticated multi-equation
approach.
TABLE 3
MIGRATION RATES AND CHANGE VARIABLES
a) Correlations (All measures refer to 1971-1976)
Out Migration Rate In Higration Rate Net Migration Rate
Growth in Employment .163 .612 . 675
Growth in the Employment Ratio -.131 -.203 -.169
Growth in Wages -.155 -.031 .070
Growth in Per Capita Income -.112 -.174 -.145
b) Regressions (Beta Coefficients) I
I-' N
Out Migration Rate Employment Wage Log Primary French Age 15-34 (1971-1976) Growth Growth Population Employment
R2=.618 .199 -.177 -.437 . 243 -.629 .508
In Migration Rate Employment Wage Log Wage Employment Mining, (1971-1976) Growth Growth Population Level Ratio Manufacturing
R2=.685 .525 .056 -.416 .245 .361 -.214
Net Migration Rate Employment Wage Log Primary Wage Employment Temperature (1971-,;(>1976) Growth Growth Population Employment Level Ratio
R2=.670 .673 .055 -.156 -.105 -.158 .271 .370
- 13 -
One other approach is worth pursuing, however, and that is to
examine the patterns of change over time in the migration rates them
selves. Table 4 presents correlations between the change ratios directly,
as well as between these ratios and other important variables. The changes
were also mapped although only one figure is presented here (Figure 4).
In each case the change in rate is measured simply as the difference
between the 1976 rate and the 1971 rate. Thus:
Net Migration Rate, 1976 - Net Migration Rate, 1971 (4)
Table 4 reveals that changes in in and out migration rates are (weakly)
negatively correlated, suggesting a kind of push-pull relationship. The
map shows that increases in the out migration rate are strongest in
Southern Ontario, particularly in the Toronto urban field -which has
been strongly urbanized during this period - and B.c., plus some mining
areas such as Sudbury and Thompson. Significant declines in out migration
rates have occurred throughout the Atlantic region· and particularly in the
wheat producing areas of the Eastern Prairies. The correlations suggest
that relatively little change is linked to social characteristics, but
that economic phenomena such as per capita income (+), level of employ
ment (+), and agricultural specialization (-) are significant. The
signs suggest that the change in out migration tends to weaken the
relationships noted in Table 2, and reflects the significant redistribu
tion of growth towards smaller, rural, places. One significant relation-
- 14 -
TABLE 4
CHANGES IN MIGRATION RATES AND OTHER VARIABLES (Correlation Coefficients)
1. Change in Out Migration Rate
2. Change in In Migration Rate
3. Change in Net Migration Rate
4. Change in Gross Migration Rate
Log10
Population, 1971
Imcome Per Capita, 1971
Employment Ratio, 1971
Age 15-34
French
Age of Settlement
Temperature
Per Cent Agricultural Employment
Per Cent Non Agricultural Primary
Per Cent Manufacturing
Employment Growth
Employment Ratio Growth
Wage Growth
Per Capita Income Growth
Out Migration, 1966-1971
In Migration, 1966-1971
Net Migration, 1966-1971
Gross Migration, 1966-1971
Mean Standard Deviation Coefficient of Variation
*Absolute value
1
1.000
-.279
-.626
.476
.115
.435
.208
.299
-.022
.065
.154
-.315
-.008
.199
.223
-.174
-.260
-.347
-.228
.360
.605
.149
4.5% 16.5
3.7
2
1.000
.814
.685
-. 216
-.260
-.094
-.256
-.027
. 010
-.006
.185
-.159
-.144
.444
.218
• 230
.337
.056
-.003
-.046
.023
14.0% 22.9 1.64
3
1.000
.312
-.204
-.295
-.075
-.458
-.084
-.020
.028
.335
-.206
-.118
.312
.198
.246
.362
.179
-.050
-.196
- .043
3.5* 4.7 1.34
4
1.000
-.115
.100
.072
-.018
-.027
.117
.173
-.097
-.181
.101
.588
.051
.015
.017
-.144
.266
.429
.120
8.0 12.3 1.54
- 15 -
ship is the temporal link between out migration and earlier in migration
and, in particular, net migration. Although out migration is basically
stable over time (remember that the correlation over the two time periods
was 0.871) it does respond to previous patterns of migration.
The map of change in in migration is more complex - less regional,
but clearly hierarchical (r .=.923). The largest declines occur in ~s
the largest cities. The greatest increases occur in the smaller centres
at the periphery of the urban fields. The correlation matrix picks out
the effect of population size(-), income per capita(-), and growth in
employment(+). A notable difference from the out ~igration column is
the lack of relationship between change in level of in migration and the
migration rates in the initial period.
The pattern of net migration (r . =.756) change (Figure 4) merges ~- tz
the two previous patterns, as one might expect. The decline in net mi-
gration to the largest places, and the increase in smaller places, is
coupled with a widespread improvement in the Atlantic Provinces and the
West, in particular the wheat belt. Declines also occur in various mining
centres. The correlations identify city size (-), per capita income (-),
age structure(-), agricultural specialization(+) and employment growth
(+) as important. The cultural environmental measures contribute very
little, and the lagged response to earlier migration rates is weak.
Changes in gross migration rates (r =.956) in contrast, have ~·~
little identifiable spatial pattern, although they are associated with
growth in economic activity, and previous levels of in migration.
The data in Table 4 suggest at least three different approaches
to explaining changes in migration behaviour. At the simplest level
I I
; I ) '\ \ I "" '; '--!.
',, /",
I ' I
I
CHANGE IN NET MIGRATION RATES
(Nr.t 1 -NM1•1l
-Over 5%
- 2-5% 0-2%
-1.5-0%
CJ Less than -1.5
-r--- __ _ 1 -r---
1 I I I
I I I I
I I I I I I I I
I I
Figure 4: Changes in Net Migration Rate
0 100 200 300 400 500 MilES .~
0 iOO ZOO :SOO 400 &00 KllOIIIIETRES
f-' 0'>
- 17 -
the presence of previous migrants in a population suggests a greater pro
pensity to move in the future. This behavioural relationship, labelled
the Axiom of Cumulative Inertia (Myers, et al. 1965) is clearly demonstrated
by strong correlations between current out-movement and previous in-
movement.
The second and third lines of argument tie back to the discus-
sion of cross sectional measures and measures of change noted earlier.
It is possible to identify, both in Table 2 and in Table 4, a change in
the response of individuals to certain kinds of (relatively) unchanging
environments. Clearly the desire to live in large cities has declined,
as has the preference for high income regions, while agricultural areas
(but not other primary producing regions) are regarded as more desirable.
At the same time the attraction of such attributes as French language,
temperature, manufacturing specialization and age of settlement appears
to be relatively constant.
The alternative -point of view interprets these shifts, not as
changes in attitude, but as responses to short-term changes in certain
volatile economic conditions. Thus employment growth or per capita in
come growth are the causal factors, which are superimposed on the basic
cultural, physical and economic structure (and thus fundamental migration
relationships) of the country. These short-run variables pose certain
operational difficulties in that they cannot be measured over appropriate
time periods. Ideally one would compare midpoints of the two time inter
vals, and possibly the average of end points would do. In fact, informa
tion is only available for the two end points, 1971 and 1976, further
- 18 -
weakening any causal arguments based on economic change as an independent
variable.
Nonetheless Table 5 does provide some support for each of the three
approaches to migration change. Let us begin, though, by reiterating the
fundamental characteristic of Canadian migration patterns. The enormous vari
ation in city size and accessibility within Canada, plus the contrasts
in culture, economic base and level of income,generate strong, continuing
patterns of migration rates - the two migration regimes noted earlier,
the gradual shifts from East to West, and from older settlements to the fron
tier. The stability of these patterns is supported by the high inter
temporal correlations. On top of this long term pattern the three kinds
of change are superimposed. First, the cumulative effect of previous
migration patterns alters the characteristics of present residents, hence
their potential to migrate. Out migration rates are clearly sensitive
to their effect, which is part of Vanderkamp's (1968) return migration.
Second, the growth of employment, and to a lesser extent increased wage
levels (but not per capita income~ are most significant in increasing
in migration rates, and also show up in the net migration change model.
Third, the shifts in the response of individuals to environmental char
acteristics show up in each model, particularly the decline in the prefer
ence for larger places relative to small rural regions. In total, the
empirical evidence supports Wolpert's (1965) behavioural model: out migra
tion is sensitive to demography and past migration behaviour; but once
the decision to move is reached in migration rates can be affected by
economic opportunity.
TABLE 5
REGRESSION MODELS OF CHANGES IN MIGRATION RATES (Beta Coefficients)
Change in Out Migration Rate
R2=.523
Employment Growth
-.167
Wage Growth
-.156
Change in In Migration Rate
R2=.339
Employment Growth
.465
Change in Net Migration Rate
R2=.342
Employment Growth
.306
Change in Gross Migration Rate
R2=.388
Employment Growth
.520
Agricultural Employment
.145
Wage Growth
.180
Out Migration Rate (t1 )
-.654
Per Capita Income
-.172
In Migration Rate (t
1)
.773
Log Population
-.052
Wage Growth
Per Capita Income
Agricultural Employment
.281
Wage Growth
-.010
-.113 .307
Out Migration Rate (t
1)
-.258
Per Capita Income
.103
Employment Ratio
-.030
Age 15-34
.335
Agricultural Employment
.191
Log Population
-.080
In Migration Rate (t1 )
.169
Non Agricultural Primary
-.121
I-' 1.0
- 20 -
Net Immigration and Natural Increase.
Given the net internal migration figures for each urban region
and estimates of net immigration, it is possible, as noted earlier, to
estimate natural increase as a residual in order to carry out a parallel
analysis of patterns of natural increase, net immigration and their
respective changes. From Table 6 it is apparent that the underlying
stability revealed in the migration rates carries over for each of these
other sources of population growth. Natural increase is correlated with
the pattern of five years previous at r=0.658 and net immigration at 0.974.
The variance in the former has increased while in the latter it has declined.
The table also confirms that the covariation among the various rates of
migration and natural increase continues, although slightly weakened: for
instance, the correlation between net immigrationand net internal migration
has declined from 0.500 to 0.408. Part of the explanation of this latter
change lies in the differing responses to population size. The correlation
of the latter with net migration has become more negative (r= -0.148),
but the correlation with immigration remains positive (r=O.l77).
The shifts in the response of natural increase and net migration to
changes in various social and economic characteristics of the Canadian
urban system are described in Table 7, which compares a series of simple
regression models developed in Part II, for the two time periods. The
relative importance of the various explanatory variables remains the same
for both dependent variables, but there are some interesting shifts.
The rate of net immigration in an urban region now responds more weakly
to the economic incentives of wages and employment level in each region,
but the relationship with the location of previous immigrants has increased.
As Taylor and Timonin (1979) point out, Canadian immigration policy has
- 21 -
TABLE 6
CORRELATIONS WITH NATURAL INCREASE AND
Variable
1.
2.
3.
4.
5.
6.
7.
8.
9.
1 2 3 4 5
Out Migration Rate, 1976 1.000
In Migration Rate, 1976 0.639 1.000
Net Migration Rate, 1976 0.114 0.837 1.000
Natural Increase Rate, 1976** 0. 314 0.117-0.073 1.000
Net Immigration Rate, 1976* 0.323 0.495 0.408 -0.072 1.000
Natural Increase Rate, 1976** 0.361 0.032 -0.216 0.658 -0.013
Net Immigration Rate, 1971 o. 366 0.478 0.354 -0.065
Population Growth Rate, 1971-1976 0.295 0.826 0. 85 7 0.332
LoglO Population 1971 -0.435 -0.354 -0.148 0.020
Mean 14.8 16.2 4.8*** 4.9
Standard Deviation 5.0 9.1 7.1 3.5
Coefficient of Variation
0.338 0.562 1.48 0. 71
0.974
0.567
0.177
0.79
2.37
3.0
IMMIGRATION
6 7 8 9
1.000
0.016 1.000
0.093 0.520 1.000
0.004 0.145 -.043 1.000
5.6 0.64 7.0 4.969
2.9 3.32 8.8 0.405
0.52 5.2 1.26 0. 082
*Defined as Immigrants from Abroad - Emmigrants (as estimated in equation 2) **Residual of Growth Net Migration - Net Immigration ***Absolute Value
- 22 -
TABLE 7
COMPARING REGRESSION MODELS OF NATURAL INCREASE AND NET IMMIGRATION
(Beta Coefficients)
Net Immigration Rate 1966-1971
Wage 0.428
Employment Rate 0.199
Log10 Population 0.114
Proportion Born Abroad 0.203
R2
0.437
Natural Increase Rate
Fish, Mines, and Forest
Proportion Age 15-34
Proportion French
R2
0.241
0. 723
-0.396
0.565
1971-1976
0.333
0.116
-0.048
0.384
0.478
0.108
0.566
-0.239
0.296
- 23 -
increased the proportion of family members entering the country and
reduced the number of independent workers, thus increasing their
temporal stablility and predictability. Both these sources of population
growth are now more social and cultural processes than economic.
The rate of natural increase, in contrast, has retained very similar
causal relationships: age structure remains the main positive factor,
which is modified negatively in French-speaking areas and increased in
primary producing areas - particularly the North - where there are large
native populations.
Keeping in mind the strong temporal stability of both natural increase
and net migration, it is possible to describe the patterns of change
(Tables 8 and 9). Table 8 suggests that the changes in rates appear to be
more responsive to economic signals than the rates themselves. Thus net
immigration is negatively linked to income and employment levels but
shifts towards higher income. On the other hand it is curiously unres
ponsive to growth in employment. Table 9 examines the correlates of the
rate change patterns. The decline in natural increase is pervasive
across the country, affecting all kinds of places in the same fashion.
(A parallel exercise, which investigated changes in household size
across the Toronto Census Metropolitan Area also found that the changes
cut across all ethnic and social class groups). Under the circumstances,
the exploration of regression models of change in natural increase seems
pointless. The one interesting (albeit weak) relationship links natural
increase growth to previous net migration levels.
The change in net immigration is more focussed, in contrast, and
appears to follow (slowly) the economic shifts of the country. Note the
shift towards agricultural, low income areas, particularly those low
turnover locations identified in Part I as "East of Montreal."
- 24 -
TABLE 8
CORRELATIONS OF NATURAL INCREASE AND NET IMMIGRATION WITH HEASURES OF ECONOMIC GROWTH
NI N Imm Growth in Growth in (1971-1976) (1971-1976) NI N Imm
1. Growth in Wage Rates 0.118 -0.112 0.039 0.248
2. Growth in Per Capita Income --0.09 3 -0.204 -0.114 0.421
3. Grm-1th in Employment 0.273 0.267 0.282 -0.108
4. Growth in Employment Ratio +0.041 -0.363 0.110 0.303
5. Growth in Out Migration Rate 0. 397 0.344 0.284 0.370
6. Growth in In Migration Rate -0.0104 0.007 0.078 0.121
7. Growth in Net Migration -0.291 Rate -0.178 -0.093 0.290
Hean 4.9 0.079 -0.71 0.159
Standard Deviation 3.5 2.37 2.76 1.146
Coefficient of Variation 0. 71 3.00 -3.89 7.21
NI = natural increase
N Imm = net immigration
- 25 -
TABLE 9
CHANGES IN NATURAL INCREASE AND NET IMMIGRATION
a) Correlations (1971 values)
Log10
Population
Income Per Capita
Employment Ratio
Age of Settlement
Temperature
Proportion French
Proportion Born Abroad
Manufacturing
Agriculture
Primary, Non Agricultural
Out Migration, 1966-1971
In Migration, 1966-1971
Net Migration, 1966-1971
Natural Increase, 1966-1971
Net Immigration, 1966-1971
Change in Natural Increase
0.022
- 0. Old
-0.061
-0.008
-0.036
0.100
-0.062
0.031
-0.057
-0.104
-0.112
0.107
0.214
-0.177
-0.102
b) Regression Models (Beta Coefficients)
Growth of
Change in Net Immigration
-0.055
-0.598
-0.414
0.079
-0.202
0.296
-0.408
-0.141
0.234
-0.083
-0.223
-0.495
-0.426
-0.072
-0.886
Net Immigration f (Income per Capita, (-0.557)
Employment Growth (-0.099),
Employment Ratio, Born Abroad, (-0.096) (0.130)
R2
= 0.379 Income Growth, Growth in Employment Rate)
(0.010) (0.152)
- 26 -
To summarize this section and point to the discussions later on, we
note the overall stability of all three sources of population growth,
with a marked decline in level of natural increase and an increase in the
level of net internal migration. Both net migration and net immigration
have responded to the striking redistribution of economic growth in this
period, which has taken place in the low income, primary-oriented regions.
MIGRATION FLOWS
Even within the relative stability of patterns of in and out migration
rates noted in the section above it is possible to have considerable change
in the patterns of place to place linkages, as long as changes in one
link are balanced by changes in others. This section examines the flows
among the 15252 possible paths linking the 124 urban centred regions,
first cartographically, and then by comparing simple regression models
of the flow over time.
An initial appraisal by means of the correlation matrix in Table 10
suggests that considerable stability also exists in the overall pattern
of contact in the system. The correlation between the log of flows for
the two time periods is 0.842. The pattern of net migration flows has
changed rather radically, however, and the correlation between 1966-1971
and 1971-1976 is only 0.384. Remember that only 350»000 (or 13.8 per cent)
of the 3,250,000 inter city migrants actually result in population change
(i.e. are recorded as net migrants). This is a modest increase from the
300,000 (13.4% of total moves) in 1966-1971, however, and suggests that
the pace of population redistribution has increased slightly.
- 27 -
TABLE 10
CORRELATIONS AMONG FLOW MEASURES OVER THfE
Variable 1 2 3 4 5 6 7 8 9
1. LoglO Flow (197 6) 1.000
2. LoglO Flow (1971) .842 1.000
3. Net Flow (1976)~~ .228 .073 1.000
4. Net Flow (1971)* .065 .233 .348 1.000
5. Flow Ratio (1976) .428 .399 .080 .022 1.000
6. Flow Ratio (1971) .300 .328 .036 .162 .464 1.000
7. LoglO Population .386 .379 -.062 -.041 -.005 -.101 1.000 Origin
8. LoglO Population .364 .375 -.062 .041 -.001 .254 -.008 1.000 Destination
9. LoglO Distance -.447 -.462 .000 .000 -.450 -.291 -.062 -.062 1.000
* Net Flow is defined as sign x (Log10 Absolute Value)
//~ ' J
' ' / ' ,' J / <
,' ? /
' I /
' \ I_ I
( )
' ' I ' " I !_,-)', '
' ' \ l, '" I
" '--I '-..... .._/
' "-, )~ ',-.1. \ '1',
, I --
Size Migrant Flow 1971-1976
- 30,000 persons
-- 10,000 persons
3,000 persons
I I
1 I
I I
I
-----~,.------r-----
I I I I
I I I I I I
I I I I I
FIGURE 5: Size Migrant Flows, 1971-1976.
/~
100 200 500 400 500 MILES
0 100 200 500 400 100 KILOMETfii.S
N 00
- 29 -
/ /
/ ---
I I
I
I
I I
I I
I
I I I I I I I I I I I I I I I I I I I I I
I I
.. \D
- 30 -
; i
0 0 ~0> ~ ~
e e 6 6 5 8 ° 8 0 8 0 g-.o..:
- 31 -
The Flow Patterns
Part I of the 1966-1971 studies presented three different maps of
flow patterns. The largest flow identifies the main migration linkages
in the urban system; the largest outflow maps the most important out
migration link for each centre; and the net flow shows population shifts
Figures 5, 6, and 7 reproduce the equivalent maps for the 1971-1976 period.
The map of the major migration flows (Figure 5) is ~s~entially the
same as the corresponding figure five years earlier. The flows largely
reflect size and distance effects which generate important migration
fields around Toronto and Montreal, but they also suggest the relatively
high rate of integration among the cities of Western Canada and the low
level of migration within the Atlantic provinces.
Figure 6 identifies the most probable destination of out movers
from each of the 124 centres. The pattern approximates the relationships
of an urban hierarchy (see Figure 1), but also incorporates the language
barrier around French-speaking Canada. It can be argued that this map
identifies the main channels of social integration in the country. As
such, it is of interest to note 15 changes in the pattern over the inter
vening five years, almost all of them tending to reduce the degree of
polarization towards the very largest places. For instance, Toronto had
a net loss of five largest links, Montreal lost four, and Vancouver lost two.
The most dramatic changes occur in the map of net migration (Figure
7) • A striking reversal of flmvs occurred between Toronto and cities
in the Atlantic provinces, while the net movements out to the peripheries
of the urban fields around Toronto, :tviontreal and Vancouver were acceler-
a ted. This is our first evidence of any substantial reordering of migra-
tion relationships from one time period to the next. It now remains to
identify which kinds of links are most affected.
- 32 -
The Relationships
In the Part II research paper the determinants of migration relation
ships were evaluated in a three step sequence: first, the overall pattern
of flows was incorporated in a gravity-type model; then the flows were
scaled by size of origin and destination to produce flow ratios, which were
then correlated with various kinds of symmetrical barriers. Finally, the
pattern of net migration flows was correlated with differences in the
characteristics of the centres at each end of the link.
The results of the experimentation with various regression models are
replicated in Table 11. (Appendix C reports on some experiments in which
the same models are applied to different subsets of observations.) The
results of the gravity formulation are virtually the same over the two
time periods, in both variance explained and in values of beta coefficients.
Less than half the variance is captured by the enormous distance and size
variations within the Canadian urban system. The flow ratios scale down
the flows by dividing by the sum of out migrants from the origin and immi
grants to the destination. With much of the size variation out of the way,
the role of distance, contiguity, hierarchy and various social and physical
barriers can be evaluated. The most notable changes are the increased
effect of distance (in the form of contiguity) over size effects (repre
sented by the hierarchy). Migrants are now more likely to move to adja
cent places of the same size than to larger or smaller places. Note also
a decline in migration across Quebec: ie. between Ontario and the Atlantic
provinces; and between Quebec and the rest of the country (the language
barrier). The net effect of these changes is a striking increase in the
effectiveness of the flow ratio model.
- 33 -
TABLE 11
cmiP ARTNG FLOW MATRIX MODELS OVER TINE (Beta Coefficient)
Log10
Flows
LoglO (fij)
Flow Ratios (f .. /'i; f .. 4:f •. )
1] 1 1JJ 1J
Origin Log
10 Population
L Destination oglO Population
Log10
Distance
Contiguity
Hierarchy
Water Barriers
Canadian Shield
Great Divide
Quebec
The Language Barrier
R2
Log10
Net Migration
sign (log10 jf .. -f .. I) 1] J1
Differences in Wage
Differences in Employment Level
Differences in Temperature
R2
1966-71
0.354
0.350
-0.414
0.457
0.262
0.251
-0.002
-0.074
-0.020
-0.031
-0.072
0.220
0.114
0.179
0.215
0.107
1971-76
0.369
0.351
-0.402
0.455
0.529
0.128
0.024
-0.109
-0.041
-0.074
-0.130
0.438
0.121
-0.016
0.106
0.026
- 34 -
Modelling net migration flows, in contrast, appears to be increas-
ingly futile. The variable, differences in the level of employment,
actually has the wrong sign in the 1971-1976 analysis, and with the focus
of high migration shifting to Alberta from B.C., the temperature diff-
erence is less relevant. Appendix C demonstrates that extracting regional
or major flow links does not really improve the model either.
As in the analysis of migration rates, it may be necessary to look
to measures of economic change to explain the observed net migration flows.
For the period 1971-1976, it is possible to derive, for those 454 flows
greater than 1000 persons (in 1966-1971), measures of employment growth,
change in employment ratios and changes in wage levels and per capita
income (see Table 12). The results indicate clearly that difference
in employment growth provide the strongest economic incentive for net
migrants (whoever they are). Improvements in the employment ratio
(i.e. proportion of population over 15 who are employed) actually
reduce the net migration flow, and higher levels of income/capita have
no effect. Undoubtedly this analysis needs a better specification in
order to better identify causal relations.
Changes in Flows
It is also possible to compare the migration periods for the two
time periods directly. At 1 a very ear y stage in the analysis the data
were adjusted to compensate for changes in the boundaries of the study
areas. Differentials in the propensity to migrate in or out of each
urban area were then adjusted for by comparing flow ratios for the two
time periods:
f. •t I E f I II f I I f lJ i ijt j ijt ijt-1 i ijt-1 3 fijt-1
- 35 -
TABLE 12
NET MIGRATION FLOWS AND DIFFERENCES IN CHANGES IN ECONOMIC CONDITIONS (Correlation Coefficients)
1 2 3 4 5
1. Net Migration~ 1971-1976 1.000
2. Change in Employment 0.466 1.000
3. Change in Employment Ratio -0.041 0.348 1.000
4. Change in Wage 0.130 0.114 -0.026 1.000
5. Change in Per Capita Income -:...o. 019 0.101 0.352 0.305 1.000
n=454 links with flows (i,j) > 1000, 1966-1971
*sign (Log10
(net migration))
Net Migration = f (Employment Growth Difference, Change in Employment Rate, (0.547) (-0.234)
Growth in Income/capita) (0 .008)
- 36 -
A value greater than one suggests that migration has a greater propensity
to occur along the link, a value less than one indicates a relative
decline in importance. Ratios and changes in flows (f .. ), net l.J
flows (NMt - NMt-l) and gross flows were also examined, but conveyed no
clear pattern.
When the changes in flow ratios for the 454 largest .l-inks (those
with flows greater than 1000 in 1966-1971) are plotted, a striking pattern
emerges. Virtually all linkages going up the urban hierarchy - from smaller
to larger places among the channels identified in Figure 1 have declined
in importance. All the linkages from larger to smaller centres, in
contrast, have become relatively larger, while flows among centres of
the same size remain roughly the same. This pattern holds throughout
the country, from St. John's to Edmonton. Many different regional
explanations can be put forth, but a relationship maintained across
such different economic and growth environments suggests a general
behavioural adjustment in the attitudes towards small centres or an
improvement in the services or quality of life to be found there.
Berry (1976) has documented similar responses occurring throughout the
Western World.
At the same time this change does not show up in changes in
migration rates. Upon reflection it is apparent that only the end points
of the hierarchy will actually be affected, since intermediate centres
will both gain from higher order centres and lose to lower order places.
The pattern of change in the flow ratios was further explored
by means of correlations and simple regressions (Table 13).
None of the correlutionn in very larbe, but the nhift in flow ratios
tends to reverse (slightly) the direction of net flows in the first
time period and also to deny the logic of economic differentials.
- 37 -
The regression models confirm that flows from larger centres have
increased, while flows from smaller centres have declined, and this
is particularly notable when relationships among the larger centres are
examined. Distance is essentially neutral in the overall model but
strongly negative in its effect on the subsample1 i.e. growth has occurred
primarily in the close together places. Most of the symmetrical barriers
used to explain flow ratios in the cross-sectional model are also neutral.
The measures of change are more interesting for they suggest shifts in
preference--towards smaller, poorer, and colder locations.
The subsample that isolates links among the larger places demon
strates even clearer, stronger patterns: declining flows are away from
larger places towards smaller places, particularly those that are near
by. This suggests that a major component of change is the expanded out
migration from the metropolitan core to the urban field, which now
appears to extend 150 to 200 miles around Montreal, Toronto and
Vancouver.
The migration patterns, while essentially stable, have also demon
strated some modest but consistent patterns of change over the last
five years. It now remains to evaluate these results in the broader
context of population growth, particularly in the light of changes in
natural increase and net immigration.
IMPLICATIONS FOR POPULATION GROWTH
The sections above show that patterns of migration rates and
flows are basically quite stable through time, and that those alterations
which do occur appear to derive from at least three different causes: a
lagged response to previous migration behaviour, short-term economic
~ 38 -
TABLE 13
REGRESSION MODELS OF CHANGES IN FLOW RATIOS
(Beta Coefficients)
FR/FR t t-1.
Net flowt -net flow 1 t-
Ail-Observations---- Subsample All Observa~ions
Log10 Population Origin
Log10 Population Destination
Log10 Distance R2
Contiguity
Hierarchy
Water Barrier
Canadian Shield
Great Divide
Quebec
Language Barrier R2
Up the Hierarchy
Difference in Employment Level
Wage Difference
Temperature Difference R2
Difference in Factor One (French to English)
Difference in Factor Two (Core to Frontier)
Difference in Factor Three (Poor to Wealthy)
Difference in Factor Four (Small to Large)
0.348
-0.100
0.002
.132
-0.013
0.040
0.012
-0.014
-0.005
-0.006
-0.028
0.003
-0.017
-0.113
-0.023
-0.044
0.018
-0.019
-0.002
-0.193
-0.229
0.090
0.508
-0.409
-0.330
0.482
-0.060
0.110
-0.022
-0.071
-0.046
0.004
-0.015
0.023
-0.134
-0.378
-0.081
-0.218
0.268
-0.121
-0.092
-0.207
-0.474
0.434
o.oe4 -0.084
0.014
-0.165
-0.028
-0.035
-0.031
0.032
0.014
0.001
-0.105
-0.095
0.020
- 39 -
growth, and an altered attitude to a particular environmental condition
by the population as a whole.
Whatever the reasons for changing migration patterns, it is
appropriate to examine them i~ the light of the most practical concern -
their implications for population growth. This section examines the
results of growth due to migration in various subsystems and levels of
the urban system hierarchy, and then compares it to the other components
of population growth: natural increase and net immigration. Finally
the size, regularity and predictability of these growth components are
discussed from the point of view of population forecasting.
Migration and Population Growth
How does the movement of over three million people from place to
place redistribute the population? Table 14 presents the main pattern of
variation in migration rates among region and size of place, which can be
compared with similar material for 1966-71 presented in Part I. The
pattern of out ~igration rates, as one would expect, remains largely the
same, with rates declining with size of centre and increasing as one moves
westward. There is a slight tendency towards increased rates of out
migration from the largest centres, while smaller places are essentially
unchanged. The pattern of in migration rates has changed more dramatically
with declines in the very largest centres, particularly Toronto and
Vancouver, and widespread increases in size orders one and two. In
addition, both the prairies and the Atlantic provinces have done better
in all size categories.
- 40 -
TABLE 14
INTERURBAN MOBILITY RATES BY ORDER AND REGIONa
OUT MIGRATION B.C. Prairies Ontario Quebec Atlantic Total
Order 4,5 11.1% 11.8 11.6 8.6 11.5 10.5
3 13.4 16.6 12.0 10.0 7.2 11.8
2 22.4 15.0 14.3 11.8 10.2 13.4
1 23.1 18.1 15.1 13.6 11.4 16.1
Total 15.7 14.1 12.5 10.1 10.0 12.1
IN MIGRATION
Order 4,5 11.6 13.5 8.3 8.0 12.4 9.7
3 22.3 14.6 11.9 6.5 7.6 11.7
2 33.8 13.9 16.9 14.1 10.7 15.8
1 30.8 16.2 18.0 9.1 12.8 16.1
Total 20.2 14.3 11.5 9.3 10.8 12.1
NET MIGRATION
Order 4,5 +0.5('-5.0) b 1.7(1.2) -33(-31) -0.6(-0.9) 0.9(1.5) -0.7(1.4)
3 8.9(1.8) -2.0(3.2) -0.1(-1.4) -3.5(0.4) 0.4(2.7) -0.1(0.4)
2 11.4(4.2) -1.1(4.5) 2.6(0.7) 2.3(2.1) 0.5(3.3) 2.4 (1. 8)
1 7.7(4.8) -1.9(3.3) 2.9(2.7) -4.5(0.2) 1.4(8.8) 0.0(2.2)
4.5(-1.9) 0.2(3.1) -1.0 ( -1. 6) -0.8(0.0) 0.8(2.3)
a Regions are groups of urban subsystems: Atlantic = Halifax Quebec = Quebec, Montreal, Ottawa Ontario = Toronto, Hamilton, London Prairies =Winnipeg, Calgary, Edmonton B.C. = Vancouver
b Parentheses indicate net shift i.e. NMt - NMt-l
- 41 -
The population redistribution effect is shown in net migration
portion of the table, where the magnitude of shift through time (NMt -
NM 1
) is shown in parentheses. Across the system as a whole a strong t-
shift towards lower order centres has occurred, so that the smallest
places, which used to lose population via migration~are now virtually
in balance, while the largest places have become a source of population
growth for intermediate centres. This is true mainly of Toronto,
Montreal, and Vancouver because Halifax and the Prairie metropoles are
still doing relatively well.
This little table has some powerful implications for forecasting.
The shift in net migration rates over 5 years is of the same order of
magnitude as the rates themselves, and if the pattern shown here were to
continue in the same direction for another five years social scientists
will be writing books in the 1980's which will be the mirror images of
the studies of the 1960's: about the abandonment of the metropolis,
while the small towns are being overrun.
The source of these net migration rates are further explored
in Tables 15 and 16 which examine the redistribution of population,
first among different sizes of city}and then among different urban sub-
systems across the country. In this case direct measures of the shift
in net migrants over the two time periods are incorporated into the
tables. Both tables indicate significant change in the pattern of
population redistribution.
- 42 -
TABLE 15
NET MIGRATION AMONG ORDERS IN THE URBAN HIERARCHY
Between places of
Order 1 120,8001
(Small)
Order 2 2 93,400 -19,600 3
(-2,000)
Order 3 -1,400 14,200 42,500 (11,900) (13' 300)
Order 4 8,100 -:,2,900 -19,400 161,600 (40,400) (17,100) (10,100)
Order 5 13,100 61,700 28,100 44,300 29,600 (Large) (23,800) (38 '900) (24,900) (19,700)
Net Migration Gain 200 92,600 -4 '100 58,500 -147,200 (74,100) (71, 300) (9 '700) (-47,900) (-107,200)
From Abroad 200 20,500 31,800 177,300 227,400
Natural Increase 160,600 186,700 172,900 272' 300 164,200
Population Growth 161,000 299,800 200,600 508,100 244,400
1971 Population 3,422,800 3,977,4oo· 3,284,300 5,641,100 5,233,600
1 Diagnonal elements are gross flows within a given order.
2 A negative flow indicates that net flow is from the column order to the row order
3 The numbers below in parenthesesindicate the net change from the previous five years i.e. NM .. t - NM.. 1 lJ' lJ 't-
1-filll.JI!J
NET MIGRATION AMONG URBAN SUBSYSTEMS
1 2 3 4 5 6 7 8 9 10 11
1. Halifax (102,600)
2. Quebec 2,300 63,000 (2,000)
3. Montreal 6,300 -17,000 204,100 (9,000) (10,400)
4. Ottawa -1,700 -4,700 -13,300 28,400 (2,600) (-1,800) (-1,600)
5. Toronto 15,100 -1,700 -19,200 4,500 288,800 (29,800) (3,000) (8, 300) (1,700)
6. Hamilton 2,000 -400 -4,300 900 -15,600 24,100 (4,000) (700) (500) (1,300) ,300)
J:'-
7. London 1,300 -100 -1,600 1,900 -500 800 42,800 w
(2,900) (900) (1,700) (1,900) (2,100) (2,200)
8. -1,000 -800 -1,500 700 4,300 -700 -800 130,700 (-1,100) (-700) (-1,200) (-3,500) (-2,600) (-1,700) (-1,300)
9. -2,400 -400 -3,500 -800 -9,700 -1,600 -2,400 -19,000 16,900 (-700) (0) ( -1 ,400) (-500) (-7,000) (-1,200) (-2,000) (3,100)
10. Edmonton -2,700 -900 -2,700 -1,500 -8,500 -1,700 -1,600 -18,200 500 31,200 (-1,600) (-200) (-1,000) (-1,400) (-7,200) (-1,500) (-1,400) (-1,100) (-2,500)
11. Vancouver -4,900 -2,000 -9,800 -4,500 -24,700 -3,500 -4,000 -23,300 -8,100 -8,300 235,200 (1,400) (-100) (700) (-2,600) (-8,000) (-1,000) (-1,500) (20,700) (3,800) (11,600)
Net Migration 14,300 -31,200 -34,000 20,900 ,300 10,700 -10,600 -62,700 32,200 29,000 95,100 Gain (42, 900) (10,300) (2,300) (-2,600) (-71,800) (-3,700) (-17,900) (37,800) (10,200) (27,300) (-25,900)
Net Immigra-2,700 -17,600 33,400 16,100 210,000 24 .• 200 17,000 12,700 23,009 23,700 112,800
tion
Natural Inc- 98,400 115,200 136,200 40,800 252,800 33,100 43,009 103,200 41,400 64,800 68,500 rease
Population llS, 400 66,400 135,600 77 '800 408,500 Growth
68,000 49,400 53,200 96,600 117 '500 276,400
- 44 -
The pattern of net migration among levels in the Canadian urban
hierarchy for the period 1971-1976 shows no clear pattern. The very
largest places (Toronto and Montreal) lose heavily. The middle centres
and the smallest centres essentially stay even, and growth accrues to
orders 2 and 4. The most important net transfers take place from order 5
to the latter two. When the pattern is compared with the previous five
years pattern, however, a clear and consistent shift is identified. The
lower the order (smaller), the more dramatic the gains; the higher the
order, the greater the declines. The net outflow from the lowest centres
has been stopped~ the growth of order 4 centxes has also been slowed,
while the net out movement from Toronto and Montreal has grown rapidly.
Part of the drama stems from the increased level of inter-order redistri
bution, which more than doubles from 128,000 to 279,000.
The pattern of regional redistribution, summarized in Table 16, is much
more complex both in cross-section and in change through time. The pattern
in 1971-1976 indicates that the Winnipeg subsystem (including Manitoba
and most of Saskatchewan) is still the big loser, and Vancouver the big
winner, but that Toronto has also become a major source of migrants and
Edmonton a major destination. The Halifax, Toronto and London subsystems
have changed signs, the first from negative to positive; the latter two, the
reverse.
Again the comparisons with the 1966-1971 pattern are more interest
ing than the cross-section itself. The level of inter-subsystem redist
ribution is essentially the same as before: 202,000 as opposed to 207,000,
but the sources and destinations have become more dispersed. The most
remarkable improvements have qeen shown by two of the erstwhile losers,
- 45 -
Halifax (gain of 43,000) and Winnipeg (gain of 38,000), and the most
dramatic declines are Toronto (-72,000) and Vancouver (-26,000). While
the latter subsystem could not be expected to maintain the frenzied in
movement of the late sixties, the changing role of Toronto - and indeed
all the Ontario subsystem - is of considerable interest. The overall
pattern suggests a substantial transfer from the urban system core -
particularly the highest order component - to the periphery, especially
the agricultural portions. }fureover the magnitude of shifts is as large as
the net flows themselves.
The shifts in the net migration pattern contrast sharply with the
stability of the natural increase and net immigration patterns. The
contribution of natural increase has declined in about every instance,
but the decline occurs across all size categories. The overall level of
net immigration has changed relatively little, and the ability of the
largest centres to attract immigrants remains essentially the same.
Montreal and, particularly, Toronto, attract more immigrants than they
lose as net migrants. The reduction in natural increase leads to a wide
spread decline in the level of population growth, but the burden of re
distributing the spatial pattern of population growth rates falls on net
internal migration, as will be shown in greater detail below.
In Figure 8, the main redistributive flows within the urban system
are plotted, and can be contrasted with Figure 8 in Part I. At this scale,
no really major changes are visible. The Western part of the urban system
is essentially the same, although the flows from Toronto to the Atlantic
provinces appear to integrate the eastern half of the country more
closely.
I 2
0
3
Order
4
0
0
\..__/
........,.. 100,000 net migrants
_ _.. 30,000 net migrants - 10,000 net migrants - 3,000 net migrants
Net immigration from abroad of less than 10,000 is not plotted
8: Net 11igration among Urban Subsystems
4 3
Order
2
.j:::--0'>
- 47 -
The Three Sources of Population Growth
Parts I and II pointed out that net migration was only one of three
sources of population growth - although it contained the greatest spatial
variance, and thus made the greatest contribution to an explanation of
population growth rates in cross-section. This section compares the
effect of net migration with that of net immigration and natural increase,
both for 1971-76 and with the earlier patterns, so that we can begin to
appreciate how truly difficult population forecasting can be.
Table 17 aggregatffi,the three components of urban growth by region
and by order in the urban hierarchy. For the system as a whole natural
increase is the predominant source of growth at 4.6 per cent, followed
by net immigration (2.2 per cent) while the total impact of net internal
migration is zero. But for any one place or subset of places either net
migration or net immigration can be just as important as natural increase.
Moreover in a cross-sectional analysis, such as we have here, the variability
of the net migration rate from place to place makes it far and away the
most significant contributor to an explanation of variations in
population growth. We will examine the temporal stability in the section
to follow.
The net migration matrix has appeared earlier, but it is worth
commenting here on the general decline of variance in this matrix relative
to 1966-1971. Only B.C. maintains a high rate of net migration; most
other values are relatively small. The changes over the past five years
have tended to benefit the erstwhile losers at the expense of the strong.
Net immigration patterns have changed little at this level of aggregation,
with a preference for the largest cities continuing along with a regional
concentration in B.C. and Ontario. A very slight decrease in the overall
- 48 -
TABLE 17
COMPONENTS OF POPULATION GROWTH, BY ORDER AND REGION (Net Change from Previous Time Period in Parentheses)
NET MIGRATION B.C. Prairies Ontario Quebec Atlantic Total RATE
Order (4,5) 0.5(-5.0) 1.7 (1. 2) -3.3(-3.1) -0 .6( -0 .9) 0.9 (1.5) -0. 7(-1.4) (3) 8.9 (1. 8) -2.0 (3 .2) -0 .1( -1.4) -.35 (0.4) 0.4 (2.7 -0.1 (0.4) (2) 11.4 (4.2) -1.1 (4.5) 2.6 (0.7) 2.3 (2.1) 0.5 (3.3) 2.4 (1.8) (1) 7.7 (4.8) -1.9 (3. 3) 2.9 (2. 7) -4.5 (0.2) 1.4 (8.8) 0.0 (2.2)
TOTAL 4.5(-1.9) 0.2 (3 .1) -1.0( -1.6) -0.8 (0.0) 0.8 (2. 3)
NET IMMIGRATION RATE
4,5 7.1(=0.6) 3.0(+0.2) 6.4(-1.7) 1.3( -0 .3) 1.2 (0.1) 3.8(-0.6) 3 2.1 (0.2) -0.9(+0.9) 1.9(-0.4) -0.5 (0.5) 0.5 (0.5) 0.5(-0.7) 2 4.4(-2.1) -0.5(+0.6) 1.3 (0.0) -0.4 (0 .6) -0.1 (1.3) -O.l(-Q.2) 1 3.3(-0.5) -0. 7(+0.8) 0.6(-1.1) -1.1 (0.8) -0.7 (0. 7) -0.7 ( -0 .3)
TOTAL 4.7(-1.2) 1.3(+0 .5) 3.9(-1.0) 0.5 (0.2) 0.1 (0.5) 2.2(-0.1)
NATURAL INCREASE RATE
4,5 1.5(-2.8) 5.4(-1.0) 4.4(-1.6) 3.6(-1.2) 4.8(-1.2) 4.0(-1.4) 3 2.8(-0.7) 9.9 (2.6) 4.5(-0.8) 5.5(-0.1) 5.4(-2.2) 5.4(-0.4) 2 6.9(-0.7) 4.8 (0.0) 4.3(-0.2) 4.7(-1.4) 5.7 (0.0) 4.9(-0.5) 1 5. 2 ( -0. 7) 5.3(-0. 7) 2.8(-0.2) 6.3(-0.6) 5.0(-1.0) 5.2(-0.3)
TOTAL 3 .2(-1.5) 5. 9 ( -0. 3) 4.3(-1.0) 4.4(-0.9) 5.3(-1.0) 4.6(-0.6)
a) The base population = 1/2 (1971 population + 1976 population over 5 years) has been used as denominator.
b) Comparisons with 1966-1971 patterns of net immigration and natural increase are based on the corrected versions from Simmons (1978b).
- 49 -
level has been accompanied by a sizeable decline in the spatial
variability as the immigration pattern has dispersed following the
trend to decentralization in the rest of the economy.
The rate of natural increase has declined slightly, affecting almost
all parts of the urban system except the prairies, where it has increased
for all city sizes - particularly in the northern centres. The contrast
between the two growth regions - the Prairies and B.C. - is interesting
The latter attracts older migrants and has the lowest level of natural
increase in the country. Declines are almost universal, but are particularly
marked for the largest cities.
The final comparison of the three sources of population growth takes
the form of a path analysis (Figure 9), which examines the contribution
of each component. This diagram replicates the analysis for 1966-1971
which is presented in Part II (as Figure 5). Note that the relative
contribution of each component to the pattern (not magnitude) of population
growth rates have not changed greatly. In the previous five year period
the corresponding coefficients were net migration (0.712), natural
increase (0.305) and net immigration (0.375). The spatial variance is
now larger for natural increase and smaller for net immigration. The
double arrows indicate the unexplained variance in each component, using
the independent variables listed at left. In most cases (all but natural
increase) the level of explanation is higher now than five years ago,
2 with the greatest improvement occurring in net immigration (R
76 0.567,
2 R 71 = 0.298). Both economic and cultural variables contribute to this
explanation.
By and large it is less easy to generalize about the 1971-1976
pattern. All kinds of variables seem to contribute to each component of
growth but the cultural measures - age of settlement, language, temperature
- 50 -
FIGURE 9
Increase
Net Migration
Net
~0.433
Path coefficients are beta coefficients
- 51 ~
and city size are consistently effective. Wage levels, in contrast,
do not contribute at all.
As an experiment Figure 10 was devised to try and sort out changes
in the pattern of population growth. In each case change is defined as the
difference in rate of migration (or natural increase or whatever) between
the two time periods. Thus,
(~) p t
Each of the measures on the right hand side of the equation was
against the independent variables from Figure 9, as well as the measures
of economic change.
(5)
The results support some of the earlier findings: the relative impor-
tance of the three sources of population growth is the same in the
dynamic model as in cross-section. Remember that these patterns have not
changed very much. In general each of the change models is weaker than
the cross-sectional equivalent, with the strongest relationship coming
from the economic change variables. The growth in employment is
particularly important in altering patterns of in migration, although not
for immigration. Natural increase responds negatively to the rise in
employment rate (i.e. more women in the labour force) and positively to
job creation. The age structure variable is measured at only one time,
and is as much a result as a cause of change. The shift in immigration
is essentially a deconcentration towards smaller, poorer centres.
- 52 -
FIGURE 10
PATH ANALYSIS: CHANGE IN POPULATION GROWTH RATE:
~ 0.818
...------..... - 0 · 241 ~Natural
Age of Settlement 1 -r .. ::::.:::-.:-=..:::.=-··· Increase I : r-------1 i i
Wage ~ -0.365 _I ~~ i ?}1>. ~~~~ion ........... I ().-
........... : '/ , Age 15-34 oc:::::-~- .....:._ ·4'-' /
-o:376·~~. ! j Primary j~. 1
1 /'-~---......
1966-1971 to 1971-1976
Temperature
4Net Migration
1 f..... .Ain I I Migration : I ' ..Jr------~ 0.210
Employment Ratio
French-····-·····_: I . :X r-... I /\% '-. -0.490 ~...,;--------r
'I ~7 CJ/ ---Log Population
A Employment
~Employment Rate
o' 1
~" 7 }';~:;:.:_;'
-0.348 ~Income/Capita-------------~
A Net Immigration
~ 0.375
- 53 -
Forecasting
Bourne and Simmons (1979) have pointed out many of the reasons why a
knowledge of future population size and composition is a significant com
ponent of policy making at all levels of government, but that paper is
not very optimistic about the capacity of social scientists to provide
this material. It is worth speculating on some of the implications of the
material presented here, which represents the first appraisal of sub
provincial population growth through time.
Forecasting is concerned with three facets of a variable: its mag
nitude or significance within the process of concern; its variability in
space and time and how this affects the process; and the predictability of
the variation: are there strong and simple links with exogenous variables,
or is there simply a large random, hence unpredictable,component? In this
case the process at issue is population growth and we can examine each of
the three components: natural increase, net immigration and net internal
migration with respect to these three concerns.
First,it is necessary to differentiate between the system-wide and
within·system performance. The effect of net internal migration, of course,
always sums to zero across the system, but both natural increase and net
immigration can vary dramatically through time (Figure 11), thus altering
the total population increment and the share of the population redistri
bution in space accounted for by each component. The population increment
from each component has quite a different spatial distribution as Bourne
and Simmons (1979) demonstrate. Figure 11 also suggests that the level of
internal migration is also variable through time, though not to the same
degree. It tends to increase slightly during periods of economic growth.
The number of net moves interprovincially is much smaller, and much more
(/) z 0 (/) a: IJJ fl.
LL. 0 (/) 0 z <( (/) ::l 0 J: I-
(/) z 0 (/) a: IJJ fl.
LL. 0 (/) 0 z <( (/) ::l 0 J: I-
1-z IJJ u a: IJJ fl.
- 54 -
500
400
300
200
Deaths 100
0 400
300
200
roo
0 4
3
Percent Change in Population
2 Percent of Canadians moving interprovincially
\
o~~~~~*S~~~~~~ 1951 'sr 'ss 1
76 YEAR
Figure 11: Components of Population Gro1·1th: Canada, 1951-1976
- 55 -
irregular over time.
The point is that the first factor affecting the future distribution
of population is the level of population growth and the relative contri
bution by natural increase and by net immigration. These levels are fore
cast (eg. Statistics Canada (91-514 and 91-520)) using quite different
kinds of information from that applied to spatial components of the system.
As the level of natural increase continues to decline (from 1,097,900 in
the first period to 933,900 in the second), and the level of net immi
gration remains roughly constant (from 455,000 to 499,800), the slowly
increasing magnitude of net internal migration (from 300,300 to 349,500)
appears more and more significant. Our forecasting concern here is to allo
cate growth among urban regions, but the growth magnitudes affect the
weighting of the different patterns.
The second, and perhaps most significant, consideration in forecasting
is the pattern of variation through space and time. Table 18 and Figure 12
convey the spatial variability of the varying components. Note that during
the first time period the variability was exactly the opposite to the
magnitude: net migration, net immigration and natural increase in that
order. Thus net migration makes the greatest contribution to the overall
pattern of growth rates across the urban system, and natural increase the
least. By 1971-1976 the variation in natural increase grew, while that
of net immigration declined, to reverse the order. At the same time
the standard duration of net migration has increased. Thus in the
cross-sectional path analysis of the last section the contribution of
net migration in explaining variations in population far outweighs net
immigration and natural increase combined. Figure 12 demonstrates the
pattern of uncertainty very clearly.
- 56 -
TABLE 18
VARIATION IN COMPONENTS OF POPULATION GROWTH
Natural Increase Net Immigration Net Migration Population Growth
1966-1971
mean 5.6 0.6 -0.2(4.8)* 6.3
s.d. 2.9 3.3 6.3 9.5
c. of v. 0.52 5.50 1.51
1971-1976
mean 4.9 0.8 1.3(4 .5)* 7.0
s.d. 3.5 2.4 7.1 8.8
c. of v. 0. 71 3.00 1.26
Changes: 1966-1971 to 1971-1976
mean -0.9(2.2)* 0.2(0.9)* 1.5(3 .5)* 0.5(4.3)*
s.d. 3.4 1.1 4.7 6.0
c. of v.
*Absolute values
**e.g. N.I.t - N.I·y-l
s.d. = Standard Deviation
c. of v. = coefficient of variation
- 57 -
40
30
20 Net Migration Rate
10
0 -15 -10 -5
50
40
Natural Increase Rate
30
20
"' (J
c: 10 Q)
:::J C" Q)
tt 0
60
50
Net Immigration Rate 40
30
20
10
0 -15 -10 -5 0 5 10 15 20%
Figure 12: Distribution of Components of PopuJation Growth
- 58 -
Over time (Figure 13) the temporal stability of net immigration
shows up very clearly-followed by natural increase and net migration
in that order. In the latter case the decline of a few very large
places supports modest increases in many small centres. In terms of
forecasting,our ability to explain or predict this distribution is
clearly a central issue.
This brings us to the third forecasting consideration - the pre-
dictability of the item in question. There is no sense in trying to
comprehend a variable with a very large random element. Here we can
draw on the accumulated experiences of this report (and also Part II).
The most surprising finding has been the spatial stability of all three
components of population growth. In each case the best predictor is
the pattern of the previous time period. Correlation between the first
and second time periods is 0.658 for natural increase, in part justifying
the faith that most population forecasters place in the simple demographic
structure. Given age structure and previous fertility characteristic9,
this component is reasonably predictable. This is even more true for
net immigration (r = 0.974), although the proclivity for earlier tl t2
immigrant clusters and larger centres has declined. It is ironic that
in the 1971-1976 path analysis the variable which makes the least con-
tribution to population growth is most easily modelled itself. The
stability of net migration is intermediate to the other two (r t tl 2
The signal of change to which it is most responsive is growth in
employment.
This discussion has so far focused on all elements of the urban
0.756).
system taken at once, in an effort to identify the most critical variables
in the population forecast. It is also possible to differentiate the
- 59 -
- 60 -
regularity of population growth components among different categories of
cities. Earlier studies (cf. Simmons, 1974) have argued that larger
cities, and cities of the industrial core, exhibit less variance in urban
growth rates. This hypothesis can be evaluated for each component of
population growth as well. Table 19 presents the range of variation in
the several components of population growth within the urban system, and then
examines the temporal correlations.
During the 1971-1976 cross-section the smaller centres contain more
variation in each growth component, as hypothesized. Large centres are
more alike and more like the national mean. The regional variation tends
to increase from East to West in absolute magnitudes, but when the
standard deviation is divided by the mean to create a coefficient of
variation, the Prairies stand out as a region of great internal contrast.
The ranking of variation: net migration, natural increase, and net
immigration seems to be consistent throughout the urban system.
The temporal correlations of the second part of the Table present
some initial surprises. Small centres are more temporally consistent
with respect to both net migration and natural increase. Apparently the
higher rate of place to place variance is stable through time - i.e.
winners keep winning and losers keep losing. Net immigration is
universally highly stable through time, but the other two components
vary markedly among subsystems. Note the shifts in net migration levels
among the larger centres in Ontario, and the smaller centres in both
the Atlantic provinces and the Prairies. Rates of natural increase have
fluctuated throughout Eastern Canada, but especially in Ontario. There
is no consistency among the population components either: where one
is stable, another may be changing.
- 61 -
The forecasting solution which suggests itself is to maintain at
least two kinds of forecasting algorithm. The first would simply allo-
cate growth by component of growth to the largest centres on a kind of
shift/share basis, using the national growth pattern. If at-l'At-l
were the local and natural population rates of increment due to natural
increase in the first time period, then
a t-1
at =(-A-- x A ) x (growth index), t-1 t
(6)
where the growth index simply adjusts for the changes in size of the
initial population of the centre over the time period:
Growth Index = Population /Population 1 t t-
(7)
If a, b, c represent the three components of population growth, then
the total population increment at a given place = at + bt + ct (8).
The efficacy of this algorithm will vary for different sets of
cities as Table 19 suggests.
at least, should be varied.
Undoubtedly the net migration component,
If b is the local population increment due t
to net migration, perhaps it would be possible to forecast local employment
growth independently. Then,
E k (e - t
t EE ee)
where k = number of net migrants in the system/Et
(9)
(10)
et' Et' are the employment increments for the urban region and the nation;
and ee, EE are the total number of workers in the region and in the nation.
Perhaps the most important component of forecasting is a sense of the
variability of the outcome. As the distributions presented here indicate,
population forecasting can never be an exact science, particularly for the
small, highly specialized centres that characterize the Canadian Urban System.
- 62 -/ TABLE 19
REGULARITY OF POPULATION GROWTH, BY ORDER AND SUBSYSTEM
(a)
Population Growth B.C. Prairies Ontario Quebec Atlantic Total
Orders 3,4,5 m. 9.9 6.7 5.3 7.2
s.d. 4.7 4.8 3.1 ... 4.5
c.v. 0.47 o. 72 0.58 0.63
Orders 1,2 m. 17.2 2.7 7.9 3.5 5.8 7.0
s.d. 12.7 9.9 8.5 5.5 3.3 9.6
c.v. 1
m. 16.2 4.3 7.5 3.7 5.9 6.8
s.d. 12.2 9.3 7.4 5.3 3.0 9.2
c.v. 0.73 2.16 0.99 1.43 0.51 1.35
Orders 3,4,5 m 1.6 -0.4 -0.4 0.1
s.d. 4.2 3.8 2.4 3.5
c.v. . .. 2.63 9.50 6.00 35.00
Orders 1,2 m 8.2 -1.9 3.2 -1.2 1.4 1.6
s.d. 11.1 5.7 7.7 5.0 3.7 7.7
c.v. 1.35 3.00 2.41 4.17 2.64 2.84
m 7.9 -1.4 2.0 -1.2 1.4 0.2
s.d. 10.6 5.2 6.8 4.7 3.4 7.1
c.v. 1.34 3. 71 3.40 3.92 2.43 35.50
Orders 3,4,5 m 5.9 4.6 5.0 5.1
s.d. 4.5 1.0 2.7 2.3
c.v. 0.76 0.22 0.54 0.45
Orders 1,2 m 5.5 5.2 3.6 5.1 4.8 4.8 s.d. 3.9 5.8 2.1 3.4 3.9 3.8
c.v. o. 71 1.12 3.9 0.67 0.82 0.79
m 5.2 5.7 3.9 5.2 4.8 4.9
s.d. 3.8 5.5 L9 3.2 3.7 3.5
c.v. 0.73 0.96 0.54 0.62 0. 77 0. 71
r:J = rean; s.d. standard deviation; c. v. = coefficient of variacti.on
- 63 -
Net Immigration
Orders 3,4,5 ;-1 2.4 2.6 0.7 1.9 s.d. 2.8 2.4 1.3 2.3
c.v. 1.17 0.92 1.86 1.21 m 3.4 -0.7 1.1 -0.4 -0.5 0.5
s.d. 2.2 1.1 1.5 2.0 1.8 2.3 c. v. 0.65 1.57 1.36 5.0 3.60 4.60 m 3.6 -0.1 1.6 -0.2 0.8
s.d. 2.2 1.7 1.9 1.8 2.4 c.v. 0.61 17.00 1.19 9.00 3.00
(b) Tem:eoral Stabilitl: Correlation coefficient between 1966-1971 and 1971-1976 values (number of observations in parentheses)
PoEulation Growth B.C. Prairies Ontario Quebec Atlantic Total Rate
Orders 3,4,5 o. 710 0.422 0.811 0.569
(2) (7) (5) (10) (5) (8) (3) (25)
Orders 1,2 0. 759 0.662 0.843 0.833 0.617 0.818
(17) (16) (21) (31) (14) (99)
0.755 0.702 0.708 0.833 0.596 0.799
(19) (21) (31) (36) (17) (124)
Net Migration Rates
Orders 3,4,5 ••. 0. 84 7 0.348 0.628 0.594
Orders 1,2 0. 798 0.367 0.900 0.666 0.482 0. 773
Total 0.794 0.450 0.827 0.664 0.480 0.756
Net Immigration Rates ---
Orders 3,4,5 ... 0.999 0.970 0.975 0.968
Orders 1,2 0.964 0.946 0.966 0.972 0.949 0.976
Total 0.956 0.979 0.969 0.973 0.955 0.974
Natural Increase Rates ---
Orders 3,4,5 ... 0.828 0.478 0.362 0.625
Orders 1,2 0.603 0.819 0.398 0.577 0.596 0.667
Total 0.627 0.785 0.457 0.557 0.602 0.658
- 64 -
CONCLUSIONS
The period 1971 to 1976 was particularly notable in Canada for the
way in which long-standing trends in the spatial concentration of eco
nomic activity were reversed. A shift in the international terms of trade
towards primary commodities produced economic expansion in rural areas,
and in areas dependent on fishing, forest products and energy production.
The redistribution of income was particularly strong, but growth in em
ployment and population also occurred in the smaller, peripheral provinces.
Only Ontario and Quebec showed a pronounced slackening in their rate of
population growth. A similar pattern was noted within the larger pro
vinces, in that population growth was more widely dispersed.
Despite this strong and consistent pattern of change, the strongest
conclusion coming out of this comparison of migration patterns across two
time periods is that migration in 1971-1976 is very much like the previous
five years. Although some consistent and interesting changes were noted,
it was found in each case that the best predictor for 1971-1976 was the
1966-1971 pattern. This regularity surprised the author, who expected much
greater change, particularly in the net effects of migration. On reflect
ion, several rationales can be put forward: the Canadian urban system is
incredibly dispersed and notably diverse, producing strong variations in
the patterns of the various components of population growth. These under
lying economic, social and cultural differences do not change rapidly
through time; they tend to maintain the demographic patterns which concern us
he~e. This stability is accentuated by the slight skewness of the rates of
- 65 -
the various growth components (see Figure 12), which over-emphasizes the
extreme rates of demographic change in a small number of places. The fact
that census observation periods need not coincide with demographic cycles
also gives an impression of stability. If, for example, a marked turn
about in migration patterns took place from 1968 to 1973, the effect would
be incorporated into both the 1966-1971 and 1971-1976 data sets. But it is
also true that very real "lag effects" do exist, in which earlier migration
behaviour affects events later on.
The corollary to the observed stability is the continuing inability to
reproduce migration patterns and, in particular, change in migration pat
terns, using regression techniques. By and large those characteristics
of the 1971-1976 patterns which cannot be explained by the 1966-1971 patterns
stilL go largely unexplained.
Within the context of these general observations we can now turn to
a list of particular findings:
1. The overall magnitude of interurban mobility has remained at about
the same level, but the proportion of net migrants increased, and in parti
cular the net shift of movers among places of different size.
2. Patterns of out, in and even net migration rates are quite stable
through time, with the pattern of the previous period explaining anywhere
from 55 percent (net) to 85 percent (in) of the variation in 1971-1976.
3. The relative importance of various social and economic variables
which describe these patterns is also largely unchanged.
4. The addition of independent variables which measure economic change,
particularly the growth of jobs, proved useful in explaining net migration
rates.
5. The analysis of change in out migration rates showed a strong
- 66 -
regional pattern, with a reduction in out migration from the Atlantic
provinces and the Eastern Prairies, while out movement has increased
strongly in Southern Ontario and B.C. - scenes of high in migration during
the earlier time period.
6. Declines in in migration are just as clearly hierarchical with
every major centre showing some loss.
7. The pattern of net iTIMigration has changed very little through
time, but the shifts appear to follow the deconcentration of the national
economy to the periphery. As a result the variation in net immigration
rates has declined markedly.
8. Natural increase~ in contrast, has become more varied in space,
and is the least stable temporally of the growth components. The process
of diffusion of the fertility decline which has moved outward from the
largest centres in Canada now appears to be virtually complete. Only a
few centres in the far north remain unaffected.
9. The map of migration flows is highly stable through time; the
pattern of net flow much less so, with some surprising reversals from
Toronto towards cities in the Atlantic provinces and from each major metro
polis to nearby smaller centres.
10. The spatial distributions which explain flow patterns appear to
operate in essentially the same fashion as before, except that any ability
to explain net migration flow has disappeared.
11. When the changes in flow over time are observed, the pattern is
quite clear. Flows from smaller to larger centres have decreased in relative
terms, while flows from larger to smaller centres have increased. This pat
tern occurs for all size groups and across all parts of the country.
- 67 -
12 _ ThesP shifts alter the results of the migration interchange am~'ng
levels in the u ·:o;:m h Lerarchy tmmrds th•2 smaller vlaces, so that in 1971-
1976 the net migration gains are fairly evenly balanced.
13. The change in net interregional flows is less drarr~tic, but tends
to reduce the variance noted in the previous period. Again the peripheral
subsystem. gain ·'· r~ the. expeitY~ of the core.
14. Net internal migration introduces more variance into population
grm-rth than both net immigration and natural incr.:ase put together. It
is also prone to rapid change through time, so tha~ ·~ h - 1~ ~ecomes a central
concern in forecasting.
15. The analysis of forecasting indicates that while the stability of
the popula.U.on grm-rth components is the most important overall relationship,
there are marked differences across groups of cities. This suggests the
need for forecasticc models ;:q .. pcopriate to city type.
The provision of migration data for the 1971-1976 period permits us to
differentiate beL\veen demographic patterns \vhich maintain existing pop-
ulation patterns or change them gradually over the long run, and. those
population changes -;-rhich can be viewed as responses to short-run conditions.
This paper has emphasized the implications of the results for the art of
population forecasting, but there are other equally important research
areas to be explored; for instance, the interaction between demographic pro-
cesses, age structure and the public sector; and the interdependence among
the growth of population, employment and income.
- 68 -
APPENDIX A
THE CANADIAN URBAN SYSTEM
This paper continues the of a set of 124 extended
urban areas (see also Simmons, 1974, 1977, 1978 a,b,) which together
comprise the Canadian urban system. If we begin with the 137 urban
places defined in 1971 (see , 1976), we can extend the definition
of the urban nodes in space until they exhaust the area of the
nation (see Figure 1) • Each urban node then includes an extended area
for which it provides services, and, in turn, depends upon as an
economic base. Brandon, for instance, is closely linked to an extensive
agricultural area, the prosperity of which determines Brandon's growth.
The operational units for this spatial extension are the counties and
census divisions used by Statistics Canada, which have the advantage
of being reasonably stable over time. When the nation's 260 census
divisions are allocated to the 137 urban places - consisting of census
metropolitan areas (C.~f.A.'s), urban agglomerations (C.A.'s) and all
other cities of than 10,000 population - the result is 124
extended urban regions.
In some cases two cities occur within the same census divisions
(e.g. Chatham and Wallaceburg) and one of them, usually the smaller,
disappears in this aggregation. In other cases an isolated census
division, served by a town or city slightly smaller than the 10,000 size
threshold, is awarded urban status (e.g. Yellowknife).
The other embellishment, essential to the urban system concept,
is the imposition of a pattern of organization or linkages among cities,
which provides a basis for grouping and aggregating urban in a
systematic manner. The simplest assumption about these relationships,
- 69 -
and the one used here, is that of a nested hierarchy. Each urban
centre is linked to a single larger place, which is assumed to provide
high order services. Five levels or orders of centres have been identi
fied. Thus, for example,Summerside (1) is linked to Charlottetown (2),
which in turn is linked to Halifax (4) and then to Toronto (5) (no
level three centre in this sequence); or Chatham (2) is linked to
Windsor (3) to London (4) and to Toronto (5) (no level one centre).
This enables us to examine relationships among cities of different
levels across Canada in uniform fashion, and to define regional sub
systems according to linkages to high order places. The assumptions
of the hierarchy and the particular linkages involved are quite
arbitrary, reflecting the inadequate theoretical discussions of these
contact patterns, and the lack of data on intercity contacts in Canada.
- 70 -
APPENDIX B
DEFINITION OF VARIABLES
Rates Analysis N=l24 Urban centred Regions
1. Population Base* = (Population(t-1) +Population over 5 years(t))l2
2. Out Migrants* = Number of persons living in a different urban region 5 years later
3. In Migrants* = Number of persons moving in from another urbcn region
4. Net Migrants* = (3) - (2)
5. Gross Migrants* (3) + (2)
6. Natural Increase* = Births - Deaths:
in 1966-1971 this information is provided in Statistics Canada (91-514).
in 1971-1976 it is estimated as a residual = population growth net migrants - net immigrants.
7. Net Immigrants*= Immigrants from abroad -emigrants:
in 1966-1971 it is estimated as a residual = population growth -net migrants - natural increase
in 1966-1971 it is estimated as immigrants from abroad - emigrants where emigrants = ((net immigrants t-1 - immigrantst-1)1 Population Baset-1 *Population Baset) * (National Emigrationt+ I National emigrationt-1)
8. Out Migration Rate* = (2) I (1)
9. In Migration Rate* = (3) I (1)
10. Net Migration Rate* = (4) I (1)
11. Gross Migration Rate* = (5) I (1)
12. Natural Increase Rate* = (6) I (1)
13. Net Immigration Rate* = (7) I (1)
14. Change in Out Migration Rate = (8) t - ( 8)t-l
15. Change in In Migration Rate = (9) - (9)t-l t
16. Change in Net Migration Rate = (10) - (10) 1 t t-
- 71 -
17. Change in Gross Migration Rate = (11) - (ll) t-1 t
18. Change in Natural Increase Rate = (12) - (12)t-l t
19. Change in Net Immigration Rate = (13) t - (13) t-1
20. Population Growth* = Population(t) - Population (t-1)
21. Population Growth Rate= (20)/ Population(t-1)
22. Change in Population Growth Rate = (21) - (21) 1 t t-
23. Age of Settlement = Number of decades since central county of urban area attained a population of 10,000
24. Temperature = Average January temperature in degrees Fahrenheit
25. Abroad (1971) Proportion of population born outside Canada
26. French (1971) Proportion of population with French mother tongue
27. Age 15-34 (1971) =Proportion of population age 15 to 34 years
28. Wage (1971) = Average employment earnings for full-time workers in the year 1970,
29. Employment Ratio (1971) holding a job.
Proportion of population 15 years and over
30. Income per Capita* Population
Total income (as reported by Revenue Canada)/
31. Agricultural Employment (1971) Proportion of employment in agricultural sector
32. Manufacturing Employment (1971) Proportion of e.mployment in manufacturing sector
33. Fish, Forest and Mine Employment (1971) = Proportion of employment in hunting and fishing, mining and forest product sectors
34. Mining and Manufacturing Employment (1971) = Proportion of employment in mining and manufacturing sectors
35. Primary Employment (1971) = (31) + (33)
36. Growth in Employment (1971-1976) Employment(t-1)
Employment(t) - Employment(t-1)/
- 72 -
37. Change in Employment Ratio (1971-1976) = (29)t- (29)t-l I (29)t-l
38. Growth in Wages (1971-1976) = Average monthly·~~~ (as estimated by Statistics Canada in "Employment, Earnings and Hours Worked" (31-002) June, 1976 - Average monthly wage (June, 1971) I Average monthly wage (June, 1971). These data were obtained in unpublished form from Statistics eanada.
39. Change in Per Capita Income (1971-1976) (30) - (30) 1
I (30) 1 t t- t-
* Defined for both 1966-1971 and 1971-1976, or 1971 and 1976.
- 73 -
APPENDIX C
A NOTE ON FLOW RELATIONSHIPS FOR SUB-SAMPLES OF OBSERVATIONS
In presenting the material on the analysis of migration flows for
1966-1971 (in Research Paper No. 98), which is reviewed in this paper in
the second section, I was frequently asked two questions: "To what degree
is the low level of explanation due to the large number of zero or very
small values in a matrix of 15,252 entries?" and "How different would
the results be if the relationships were evaluated within regions - the
Atlantic Provinces, the industrial corridor and so on?" Both these problems
can be answered readily within the SPSS format.
The first exercise (Table C.l) was to isolate a subsample of the larger
flows in the matrix. All dyadic links with more than 1,000 migrants between
1966 and 1971 were selected. (This is the same procedure used by Termote
and Frechette (1979) in their study of interprovincial flows.) The table
suggests that, by and large, relationships are weaker for the subsample -
apparently the sampling and measuring error in the smaller flows does not
explain the weakness of the models. The relative values of the coefficients
are not greatly altered, however. It appears that the general structure of
relationships is stable across the size hierarchy.
The regional variations are examined by extracting only those flows
which begin and end in the same subsystem. In this case we will examine
only the 1971-1976 relationships, and for four regions: the Atlantic
Provinces, Quebec, Ontario and Western Canada. The results suggest that
the within region models usually achieve higher levels of explanation, and
this is true of each version of the model. In particular, the gravity
model performs better when it does not have to deal with interregional
- 74 -
patterns. On the other hand the models incorporating physical barriers
perform less effectively because they tend to separate the regions,
rather than subdivide them. Contiguity remains as remarkably effective.
The net migration models are more effective for subsamples, but only
marginally, and in quixotic ways. For instance, we observe net migrant
flows towards high unemployment areas, and towards low wage areas -
picking up the diffusion of population into the urban fields of the largest
metropolitan areas.
All in all, the isolation of regional subsamples does not substantially
alter the flow models developed within the text of the paper. The gravity
model performs well in all regions. The barrier effects at least show
the right signs, but the net migration relationships are still beyond
comprehension.
- 75 -
TABLE C.l
FLOW RELATIONSHIPS FOR LARGER FLOWS (Beta Coefficients)
Log10 fij Independent
Log10 Origin Population
Log10
Destination Population
Log10
Distance. R2
Flow Ratio
f .. / r. f .. 'E. f lJ c lJ i ij
Contiguity
Hierarchy
Water Barrier
Canadian Shield
Great Divide
Quebec
Language Barrier R2
Log10
Net Migration
1966-1971
All Observations
0.354
0.350
-0.414
0.457
0.262
0.251
-0.002
-0.074
-0.020
-0.031
-0.072
0.220
(sign) Log10 Absolute Value of fij -fji
Wage Difference
Employment Level Difference
Temperature Difference R2
0.114
0.179
0.215
0.107
1 Sub sample
0.502
0.536
-.335
.325
0.152
0.124
0.04 7
-0.069
-0.069
-0.024
-0.077
0.088
0.024
0.303
0.186
0.136
1971-1976
All
0.369
0.351
-0.402
0.455
0.529
0.128
0.024
-0.109
-0.041
-0.074
-0.130
0.438
0.121
-0.016
0.106
0.026
Subsample
0.571
0.400
-0.384
0.322
0.476
0.075
0.156
-0.061
-0.085
-0.011
-0.072
0.306
-0.072
0.012
0.059
0.008
1 Subsample contains only those 454 links with flows greater tpan 1000 during the 1966-1971 period.
- 76 -
TABLE C.2
FLOW RELATIONSHIPS FOR REGION FLOW GROUPS (Beta Coefficients)
All Atlantic Quebec Ontario West
Log Flows Log10
Origin Population 0.369 0.551 0.457 0.564 0.425
(LoglO ) Log10 Destination Population 0. 351 0.500 0.498 0.491 0.474
Log10
Distance -0.402 -0.707 -0.343 -0.390 -0.413 R2 0.455 0.635 0.603 0.737 0.587
Flow Ratio
(L ./ . l.J 1 i fij) Contiguity 0.529 0.596 0.594 0.679 0.545
Hierarchy 0.128 0.126 0.053 0.094 0.124
Water Barrier 0.024 -0.178 0.056 -0.039
Canadian Shield -0.109 -0.033
Great Divide -0.041 -0.001
Quebec -0.074 -0.090 -0.015
Language Barrier -0.130 -0.087 -0.128 R2 0.438 0.563 0.415 0.527 0.358
Log Net Migration Wage Difference 0.121 0.156 0.177 -0.355 -0.149
(sign) Log1E1/fij- Employment Level Difference -.016 0.352 0.334 0.002 -0.001
fij) Temperature Difference 0.106 -0.072 0.221 0.095 0.050 R2 0.026 0.120 0.189 0.136 0.031
n 15252 272 1254 930 762
- 77 -
The Flow Analysis Variables
FLOWS
GROSS FLOWS
NET FLOWS
LOG NET FLOWS
FLOW RATIO
NET FLOW RATIO
FLOW CHANGE
FLOW RATIO CHANGE
NET FLOW CHANGE
DISTANCE
CONTIGUITY
BARRIERS BD
BS
BW
BF
BFF
HIERARCHY
DIFFERENCES
Fl (Ethnicity)
F2 (Demography)
Number of Migrants from i to j, 1966 to 1971.
Migrants i to j + Migrants j to i
Migrants i to j - Migrants j to i
Sign x LoglO (Absolute Value (net flow)
Migrants i to j I~ Migrants from i xI migrants to j
(Migrants i to j - Migrants j to i) I i migrants from i x ~migrants to j
Flow I Flow 1 t t-
Flow Ratiot I Flow Ratiot-l
Net Flowt - Net Flowt-l
straight line distance in miles between city centres
Value 1 if places are adjacent
Value 1 if places are separated by the Great Divide
Value 1 if places are separated by the Canadian Shield (Thunder Bay to Sault Ste. Marie)
Value = 1 if places are separated by the Gulf of St. Lawrence or Georgia Strait
Value = 1 if English-speaking places are separated by intervening French-speaking region
Value = 1 for movement between English and Frenchspeaking urban regions
Value = 1 if a hierarchical commercial linkage exists between the two places.
Fld - Fl 0
F2d - F2 Differences in Factor Scores 0
F3 (Economic Disparity) F3d - F3 0
F4 (Metropolitanism) F4d - F4 0
CITY SIZE RATIO
WAGE DIFF.
EMPLOYMENT RATIO DIFF.
TEMPERATURE DIFF.
EMPLOYMENT GROWTH DIFF.
- 78 -
Log10
Population Destination - Log10
Population Origin
Wage at Destination - Wage at Origin
ER (Destination) - ER (Origin)
Temperature (Destination) - Temperature (Origin)
Employment Growth Rate (Destination) - Employment Growth Rate (Origin)
INCOME/CAPITA GROWTH DIFF. Income/Capita Growth Rate (Destination) -Income/ Capita Growth Rate (Origin)
- 79 -
REFERENCES
Berry, Brian J.L., ed. (1976). Urbanization and Counter Urbanization. Urban Affairs Annual Review. Beverly Hills, California: Sage Publications.
Bourne, L.S. and Simmons, J.W. 1979. "Canadian Settlement Trends: An Examination of the Spatial Pattern of Growth, 1971-1976", Report Prepared for the Ministry of State for Urban Affairs. Major Report No. 15. University of Toronto, Centre for Urban and Community Studies.
Greenwood, Michael. 1975a. "Research on Internal Migration in the United States: A Survey11
, 13, 2, pp. 397-433.
Greenwood, Michael. 1975b. "A Simultaneous Equations Model of Urban Growth and Migration", Journal of the American Statistical Association 70 (December), pp. 797-810.
Kalbach, Warren. 1970. The Impact of Immigration on Canadian Population, 1961 Census Monograph. Ottawa: Dominion Bureau of Statistics.
Muth, Richard F. 1971. "Migration: Chicken or Egg?'' Southern Economic .;:;.;.;;_;;;;..;;;..;:.:.== 37 (January), pp. 295-306.
Myers, G.C., McGinniss, R., and Ma.srick, G. 1965. "Preliminary Assessment of a Stochastic Process Model of Internal Migration",
31, pp. 256-257.
Ray, D.M., ed. 1976. Canadian Urban Trends: Vol. 1, National Perspectiv'E!. Toronto: Copp Clark.
Simmons, James W. 1974. "Growth of the Canadian Urban System~', Research Paper No. 62, University of Toronto, Centre for Urban and Community Studies.
Simmons, James W. 1977. 11Migration and the Canadian Urban System: Part I, Spatial Patterns", Research Paper No. 85. University of Toronto, Centre for Urban and Community Studies.
Simmons, James W. 1978a. ''Migration and the Canadian Urban System", Part II, Simple Relationships", Research Paper No. 98. University of Toronto, Centre for Urban and Community Studies.
Simmons, James W. 1978b. "Migration in the Canadian Urban System", in R.L. Irving, ed., Readings in Canadian Geography, 3. Toronto: Holt, Rinehart and Winston, pp. 47-67. This is a revised version of Simmons (1977), which incorporates direct estimates of natural increase.
Statistics Canada. 1974. Population Projections for Canada and the Provinces, 1972-2001. Catalog 91-514. Ottawa.
- 80 -
Statistics Canada. 1979a.Population Projections for Canada and the Provinces, 1976-2001. Catalog 91-520. Ottawa.
Statistics Canada. Quarterly Estimates of Population for Canada and the Provinces. Catalog 91-001. Ottawa.
Statistics Canada, 1979b. "Revised Estimates of Population for Counties and Census Divisions, 1973 to 1975", Catalog 91-513. Ottawa.
Stone, Leroy 0. 1969. Migration in Canada: Regional Aspects. 1961 Census Monograph. Ottawa: Dominion Bureau of Statistics.
Taylor, C.E. and Timonin, I.M. 1979. "Immigration Levels, Emigration and Canadian Population Growth: A Review", Paper presented to the Canadian Population Society, Saskatoon.
Termote M. and Frechette, R. 1979. "Les Variations du Courant Migratoire Interprovincial", Report prepared for the Ministry of State for Urban Affairs. Montreal: Institut National de la Recherche Scientifique.
Vanderkamp, John. 1968. "Interregional Mobility in Canada: A Study of the Time Pattern of Migration", Canadian Journal of Economics, 1 (August), pp. 595-608.
Wolpert, Julian. 1965. "Behavioral Aspects of the Decision to Migrate", Papers and Proceedings of the Regional Science Association, 15, pp. 159-169.